Build a World-Class Content Team with Claude Skills

How to build an AI content system using Claude Skills and Claude Code Skills.

Most AI content sounds like AI wrote it. That’s not an AI problem, it’s a workflow problem.

A large majority of marketers still struggle to use ChatGPT, Claude, or Gemini to craft engaging content. Typing “write me a LinkedIn post about [topic]” will get you something grammatically perfect, but instantly forgettable. It reads like a generic copy of every Ted Talk you’ve ever watched.

The problem isn’t AI; the problem is asking AI to do everything in one shot. You’re asking it to know your audience, match your voice, find the right angle, structure the post, and nail the opening, all from a single prompt.

The best AI marketers aren’t writing better prompts. They’re building systems. This is why I’ve been emphasising that so much of marketing’s future will be mastered by people with engineer-led mindsets. Systems builders.

This tutorial shows you how to build a content system using Claude Skills. By the end, you’ll have three installed skills that turn your existing content into structured, post-ready talking points, written in your voice, for your audience.

I’ll walk you through every skill in the content system, including the outputs of the advanced skills, so you can see exactly how the full engine works and start to understand why Claude Code skills are so powerful. Three of the skills are yours to install and use today. The rest will be available to paid subscribers soon as part of a more in-depth post on Claude Code skills.

What Are Claude Skills (And Why Should You Care)

A prompt is a conversation. A skill is a system.

When you type something into Claude, you’re having a one-off conversation. You’re continually re-explaining your audience, your voice, your context, what you want, every single time.

A Claude Skill is different. It’s a set of instructions you give Claude once, and it remembers them forever.

When I say “extract viral talking points from this transcript”, my Talking-Point Extractor skill already knows:

• Who exactly my audience is (CMOs and VP of Marketing)

• The viral talking points that work for my audience (Educational, Spicy Take, Data Nugget, Story Spark)

• The quality bar to hit

• How to deliver the output

Nothing to re-explain. Not having to constantly give the AI a role to act out, e.g., “act as a content strategist who specialises in B2B marketing.” The skill handles all of that.

Claude Code Skills take this further. They can execute code, call external APIs, and automate research across platforms. A regular Claude Skill is like a smart assistant who follows your playbook. A Claude Code Skill is like that same assistant with access to the internet, a database, and a script that runs the whole workflow for you.

In this tutorial, most of what you’ll build are Claude Skills, instructions you install in Claude.ai with a single upload. I’ll also show you Claude Code Skills that handle things regular skills can’t: scraping writing samples from LinkedIn/X, and researching what your audience is actually talking about on Reddit and X in real time.

You don’t need to know how to code for any of this.

---

The system has three phases. Each phase builds on the last.

Phase 1: Setup — Create the two assets that make everything downstream smarter: an audience profile and a writing style card that you store in memory. You build these once and reuse them across every piece of content you create.

Phase 2: Ideation — Turn raw content into structured talking points. Drop in a podcast transcript, a conference talk, a blog post, or an article, and get back post-ready ideas categorised by type.

Phase 3: Enhancement — Elevate your talking point with hooks that stop the scroll and enrichments that bring the post to life, a relevant story, a case study with real numbers, or a quote from someone your audience respects. Treat content as modular blocks.

What we’ll cover:

Setup: Content Audience Profiler → a research-backed profile of exactly who you’re writing for. Writing Style Analyzer → a reusable card that captures how you write.

Ideation: Talking-Point Extractor → post-ready ideas pulled from your existing content.

Enhancement: Hook Creator → 5 psychologically distinct openers, ranked for your content type. Post Enricher → a story, case study, and quote option mapped to your talking point.

The first three skills are yours to install today. I’ll walk you through every skill in the system, including full outputs from the advanced skills, so you can see exactly how the pipeline works end to end.

Let’s start building.

---

Section 4: Skill 1 — Content Audience Profiler

Most content fails before a single word is written.

Not because the writing is bad. Because the writer never decided who they were writing for. They write for “marketers” or “business leaders” which means they write for no one.

David Ogilvy famously kept a photo of the person he was writing to pinned above his desk. Every headline, every line of body copy, is written to that one person. Not an audience. A person.

This skill is the AI version of that photo.

The Content Audience Profiler asks you two questions:

1. Who is your target audience? (e.g., “CMOs at B2B SaaS companies”)

2. What’s your core content topic? (e.g., “AI for marketers”)

That’s it. Two inputs. From there, it runs 4-6 research searches and builds a one-page profile covering who they are, what keeps them up at night, what content they actually engage with, how they talk, and who they trust.

Here’s what the output looks like for my audience — VPs and CMOs of Marketing:

The profile surfaces things you might believe but haven’t articulated. For example: 63% of CMOs cite budget constraints as their top challenge (Gartner, 2025), but the real anxiety is deeper — they’re being asked to do transformational work with operational budgets. That’s a content angle. That’s a hook. And you’d miss it if your audience definition was just “marketing leaders.”

Once this profile exists, every other skill in the system uses it. Your talking points get tailored to these pain points. Your hooks speak to these anxieties. Your stories land because they map to situations this audience actually faces.

Build it once. Use it forever.

→ Download the Content Audience Profiler skill

---

Section 5: Skill 2 — Writing Style Analyzer

Here’s a test. Open 5 LinkedIn posts from a creator you admire.

You’ll hear patterns. The sentence length. The words they reach for. The way they open, the way they close. Their punctuation.

That’s their voice. And that’s what AI kills.

You see, every model has its own voice, not yours. Because of AI’s training set, it’s helpful, balanced, slightly formal, relentlessly thorough. Great for answering questions. But, the death of great content. Your audience follows you because of how you say things, your personality, not just what you say. I’ve referred to this before as ‘personality-led growth’. Using AI strips you of that. You sound like everyone else.

In the early 1960s, Hunter S. Thompson got a job at Time magazine. He wasn’t writing; he was a copyboy. But every night, he’d sit at his desk and type out The Great Gatsby and A Farewell to Arms, word for word. Not to plagiarise. To absorb. He wanted to feel what it was like to write at that level, the rhythm, the pacing, the way Fitzgerald built tension in a sentence.

He later told Charlie Rose: “If you type out somebody’s work, you learn a lot about it. Amazingly, it’s like music.”

This practice is called copywork. Benjamin Franklin did it. Jack London did it with Kipling. Robert Louis Stevenson did it. For centuries, the way you learned to write was by studying how great writers wrote, then developing your own voice from that foundation.

The Writing Style Analyzer is copywork for the AI age.

Feed it 5-10 writing samples from any creator, and it produces a Style Card, a document that captures exactly how that person writes. Sentence length, punctuation habits, vocabulary, structure, tone, all of it.

You can use these two ways.

Capture your own voice. Feed in your best LinkedIn posts, X posts, and Substack posts, and get a Style Card that ensures AI always sounds like you.

Study creators you admire. There’s a creator whose X threads always get engagement. Someone whose LinkedIn posts your audience loves. Feed their public content into the Analyzer, get their Style Card, then edit it to your tastes. Keep the structural patterns that work, swap in your vocabulary and tone. It’s Thompson typing out Gatsby, but instead of months of typing, it takes 2 minutes.

You can build different style cards for different platforms. Your LinkedIn voice might be different from your newsletter voice. A creator who kills it on X might have a style that doesn’t translate to long-form. One card per platform, each one tailored.

Here’s a section from my LinkedIn Style Card:

The Style card allows you to map all future content to the same style.

→ Download the Writing Style Analyzer skill

---

Section 6: Skill 3 — Talking-Point Extractor

You’re sitting on more content than you think.

That podcast you did last quarter. The conference talk. The internal strategy doc you wrote for your team. The Slack thread where you explained your position on AI agents to your VP of Demand Gen. That 45-minute customer webinar.

All of it is raw material. All of it contains ideas your audience would engage with. And almost none of it is being used. AI is your opportunity to turn this into engaging content.

The Talking-Point Extractor fixes this. Drop in any piece of existing content, a transcript, an article, a rough draft, even meeting notes, and it pulls out the most post-worthy ideas, structured and categorised.

It doesn’t just summarise. It hunts for the moments that would make someone stop scrolling.

Every talking point gets categorised into one of four types:

Educational — a lesson, framework, or how-to that your audience can learn from. The kind of post people bookmark.

Spicy Take — a provocative opinion that challenges conventional thinking. The kind of post that fills your comments section.

Data Nugget — a jaw-dropping statistic with a clear takeaway. The kind of post people screenshot and share with their team.

Story Spark — a short story with a clear moral. The kind of post that gets reshared because it makes people feel something.

These four categories aren’t arbitrary. They map to the four reasons people engage with content: to learn something, to debate something, to validate a decision, or to connect emotionally. Every viral post you’ve ever shared or engaged with fits one of these.

Here’s what the output looks like when I feed in a transcript from a recent MATG episode (just one of my talking points):

A 19-minute podcast produced multiple viral talking points.

The extractor doesn’t generate generic ideas. Every talking point includes the specific detail, the number, the example, and the quote, which makes it worth posting. If it can’t find enough substance for a talking point, it skips it. No filler. Combine with your audience profile, e.g., extract viral talking points for my [saved audience] to match them exactly to your audience.

→ Download the Talking-Point Extractor skill

---

The Full Content System

This is one system I’ve started to build out for content. However, I’m working on building systems for everything. The three skills you installed are a great foundation. They give you a research-backed audience profile, a reusable voice, and a system for turning existing content into structured ideas.

The full content system for this workflow includes four more skills, all in Claude Code:

Content Scraper (Claude Code Skill) — Automatically collects 5-20 writing samples from any creator across LinkedIn, X, and Substack. Instead of manually copying and pasting posts into the Style Analyzer, this does it in under a minute. Feed it a name, get back a clean file of their best content ready for analysis.

Viral Talking Points (Claude Code Skill) — Researches what your audience is actually discussing right now across Reddit, X, and the web. Instead of guessing what topics will resonate, you see real conversations with real engagement data — upvotes, comments, likes — and extract talking points from what’s already proven to get attention.

Hook Creator (Claude Code Skill) — Generates 5 scroll-stopping openers for any talking point, each one using a different psychological trigger: cognitive dissonance, curiosity gaps, surprise via data, narrative tension, and loss aversion. The hooks are stack-ranked based on your talking point type and adapted for your platform. It researches what hook patterns are currently performing in your niche, analysing high-engagement posts from the last 30 days on LinkedIn and X, so the hooks aren’t based on generic best practices. They’re based on what’s working right now, for your audience.

Post Enricher (Claude Code Skill) — This is my favourite Finds the story, case study, or authority quote that turns a solid post into a great one. It searches across the web for relevant stories from business, sports, science, and history, then verifies the facts, checks the attribution on quotes, and finds recent case studies with real numbers from published sources. The difference between a post people read and a post people remember is almost always one of these three things , a narrative that makes the abstract concrete, a real-world example with numbers, or a quote from someone the audience respects. This skill finds them for you instead of relying on what the AI already knows.

Here is the visual of this content system:

These are part of the Claude Code for Marketers course I’m working on, much of it I’ll share with my paid subscribers here, starting with the above 4 skills. However, don’t fear, I will start sharing Claude Code use cases for free as well :)

Until Next Time,

Happy AI’fying

Kieran